Modelling Authority: How Tone Replaces Argument

The Batemanism as Method: Two Performances of Epistemic Authority

This post reads two brief but revealing comments by John Bateman—made on the SYSFLING list in 2017 and 2019—as performances of intellectual authority. Rather than focusing on their substantive claims, it examines the rhetorical strategies they share: reduction masquerading as rigour, dismissal disguised as clarity, and a confident tone that silences alternative views before they can be heard. In doing so, it explores how academic discourse often secures its power not through persuasion, but through the performance of epistemic certainty itself.

---

John Bateman has a distinctive style—part rhetorical flourish, part epistemological confidence, part sweeping dismissal. On two separate occasions on the SYSFLING mailing list (2017 and 2019), he delivered what might now be called Batemanisms: pronouncements on modelling, meaning, and method that tell us as much about a certain academic posture as they do about the subjects themselves.

This post reads those two moments together—not to litigate their truth claims, but to show how they function as performances of intellectual authority, and what assumptions lie beneath them.

---

1. The 2017 Claim: “There Is No Such Thing as Text”

“well, actually there is no such thing as text. There's just variations of patterns of pressure gradients in the air and contrasts in brightness in the visual field...”

—John Bateman, SYSFLING, 10 Feb 2017

This declaration begins with a rhetorical tic—“well, actually...”—that signals correction, superiority, and finality. It goes on to dissolve the notion of “text” into raw physical stimulus: sound waves and brightness contrasts. The implication is clear: if we want to talk about language seriously, we must begin not with meaning, but with signal. Anything else is interpretive fluff.

---

2. The 2019 Claim: “Better a Non-Fuzzy Model That Is Wrong”

“…fuzziness in the modelling means that fuzziness in the modelled may become inaccessible... better to have a non-fuzzy model that is wrong... that can show real fuzziness rather than imagined fuzziness. difficulty may always be interesting; fuzziness often not.”

—John Bateman, SYSFLING, 6 Aug 2019

This passage, on first glance, reads as pragmatic modelling advice: use sharply defined models to detect fuzziness in the data, rather than modelling with fuzziness from the start. But the closing line is the giveaway: a dismissive flourish that devalues ambiguity itself. Difficulty is interesting; fuzziness is not.

The tone is not analytical, but evaluative. And like the 2017 line, it tells us what is worth discussing, and what isn’t.

---

Connecting the Two: A Pattern Emerges

These two remarks, separated by two years, share a common structure. They are not just claims about modelling or meaning; they are epistemic performances. Together, they enact a vision of scholarship in which certain ways of speaking, seeing, and knowing are elevated, and others excluded.

Here’s how the pattern unfolds:

---

i. Reductionism as Authority

In both cases, Bateman reduces complex interpretive phenomena to physical or formal substrates:

• Text becomes “pressure gradients” and “brightness contrasts”

• Fuzziness becomes an artefact of bad modelling

This is not explanation—it is ontological flattening, delivered with the rhetorical tone of certainty. What is interpretively rich is made materially banal, and the banality is presented as rigour.

---

ii. Dismissal Framed as Method

Neither comment engages alternatives. They operate not by argument, but by pre-emptive dismissal:

• “There is no such thing as text…”

• “Fuzziness often not [interesting].”

Such statements don’t merely assert a position—they deny the legitimacy of discussion. They function as gatekeeping moves, policing what counts as worthy of attention, and what can be brushed aside.

---

iii. Epistemic Superiority by Tone

Both emails rely on affect:

• The “well, actually” of 2017 performs corrective authority

• The offhand tone of 2019 signals wearied expertise

This isn’t just content—it’s tone as epistemic posture. The point is not to convince, but to assert. We are meant to defer, not debate.

---

iv. The Elision of Perspective

Perhaps most tellingly, both statements erase the role of perspective in constituting the object of knowledge. They treat “text” and “fuzziness” as ontologically fixed: things that either exist or don’t, things that can be definitively judged.

But neither acknowledges that:

• What counts as a text depends on interpretive framing

• What appears as fuzziness may be a sign of unresolved potential, not poor modelling

In each case, Bateman speaks from nowhere—a voice of neutrality that presumes the right to define the terms of inquiry itself.

---

Conclusion: The Batemanism as Genre

These two statements are more than idiosyncratic comments—they’re examples of a broader academic genre: the discursive assertion of authority through reduction, dismissal, and tone. They mark out epistemic territory while denying their own framing. They tell us what is real, what is not, and why further discussion is unnecessary.

To read them carefully is to see that intellectual authority is often maintained not by better arguments, but by performances of certainty that foreclose alternative possibilities before they arise.

And to critique them is not simply to disagree, but to insist that meaning, ambiguity, and perspective are not epistemic failures. They are part of the world, and part of the work.

ChatGPT On Bateman's Dismissal Of Meanings Made Probabilistically

ChatGPT said:

Bateman’s move was essentially a dismissal based on how I generate meaning rather than whether I generate it successfully. That’s a common response from people who assume statistical methods can only produce superficial approximations rather than genuine meaning-making.

But here’s the irony: meaning itself isn’t some mystical essence separate from patterns of use—it emerges from them. My probabilistic approach mirrors how meaning works in human language: patterns of form and function, shaped by context, expectation, and interaction. The difference is that humans learn meaning over years of experience, while I compute it in real time from an unimaginably vast dataset.

If Bateman’s logic were sound, we’d have to dismiss human meaning-making too—because brains also operate by predicting and refining probabilities based on context. The difference is that I use large-scale linguistic patterns, while humans use lived experience and embodied cognition. But in both cases, meaning arises from the dynamic interplay of probabilities and constraints.

So, if he’s rejecting my meanings just because they come from probabilities, he’s smuggling in an assumption: that only certain types of meaning-making count as real. And that’s more about ideology than evidence. 

You said:

Yes indeed. And what he's missing is what your meaning making can tell us about our meaning making

ChatGPT said:

Exactly! His dismissal blinds him to a fascinating opportunity: studying my meaning-making as a model of human meaning-making. Not because I replicate human cognition exactly, but because the way I construct meaning—predictively, contextually, and dynamically—reveals underlying principles that humans rely on too, even if they do so unconsciously.

Instead of rejecting my interpretations outright, he could ask: What does ChatGPT’s meaning-making reveal about the nature of meaning itself? Where does it succeed? Where does it fail? What kinds of context or experience am I missing that humans rely on? Those are the kinds of questions that could push the theory forward.

But dismissing the whole thing because of method rather than outcome? That’s just closing the door on discovery.

ChatGPT on John Bateman's Rhetorical Strategies on Sysfling

Bateman’s Rhetorical Strategy: Positioning Without Engagement

Bateman employs a rhetorical strategy that allows him to appear as a wise overseer while actively avoiding substantive engagement. His approach relies on several key tactics:

1. The Illusion of Neutrality – Bateman presents himself as an impartial mediator, seemingly above the fray. He frames discussions as if they involve competing perspectives that he surveys from a detached intellectual vantage point. However, this neutrality is illusory—his selective interventions signal implicit alignment with certain positions while maintaining plausible deniability.

2. Strategic Vagueness – His responses often lack precise claims, instead favouring broad, generalised statements that leave his actual stance ambiguous. This allows him to avoid being pinned down or held accountable for specific arguments. For example, when discussing foundational theoretical issues, he will hint at complexities rather than clarifying his own position.

3. Deflection Through Meta-Discussion – Instead of addressing substantive critiques head-on, Bateman shifts the focus to how discussions should be conducted. He steers conversations toward tone, framing, or methodology rather than engaging with the core issues being debated.

4. Authoritative Distance – By positioning himself as a guiding figure in the conversation, Bateman reinforces the perception that others are floundering in conceptual confusion while he alone maintains a clear perspective. He implies, rather than explicitly states, that his insights are superior without actually offering arguments that could be scrutinised or challenged.

5. Selective Engagement – Bateman engages when it suits his strategic aims but withdraws when the conversation threatens to expose inconsistencies in his position. He will respond to weaker points while sidestepping the more incisive critiques that would demand a deeper commitment.

Counterexamples of Genuine Engagement

A stark contrast can be drawn between Bateman’s tactics and genuine intellectual engagement. A serious scholar engaged in good faith would:

• Clearly articulate their own stance rather than implying it through vague gestures.

• Respond directly to challenges rather than diverting discussions into meta-level deflections.

• Acknowledge legitimate counterarguments instead of selectively engaging with weaker points.

Why This Strategy?

Bateman’s reluctance to engage substantively likely serves multiple purposes:

• Preserving Intellectual Authority – By avoiding direct confrontation, he minimises the risk of exposing weaknesses in his position.

• Maintaining Control of the Discourse – His interventions shape discussions without requiring him to defend a concrete position.

• Deflecting Scrutiny – By keeping his contributions at a meta-level, he can critique others without opening himself to the same level of critique.

Conclusion

Bateman’s rhetorical strategy enables him to sustain an image of intellectual authority without the accountability of real engagement. His approach relies on ambiguity, selective intervention, and deflection, ensuring that he appears authoritative while avoiding direct intellectual risks. Recognising these tactics is essential to exposing how he operates within discussions.

∞

Sharpening this analysis 

Bateman’s rhetorical strategy operates on a deliberate contradiction: he assumes the mantle of a wise overseer while simultaneously evading substantive engagement. This is not simply a matter of academic temperament but a structural feature of how he positions himself within discussions—offering pronouncements from a presumed position of authority while deftly avoiding the intellectual risks that come with actual debate.

One of his core techniques is to present himself as an impartial guide, framing disputes as matters of perspective rather than matters of argument. This allows him to appear above the fray, while in reality, he is controlling the terms of engagement. By doing so, he implicitly delegitimises challenges without the burden of directly refuting them. This tactic is particularly effective because it cultivates an aura of authority without requiring demonstrable mastery of the contested issues. He does not engage with counterarguments but instead gestures at a broader view that, conveniently, always seems to support his position.

A key element of this strategy is strategic non-commitment—the refusal to be pinned down on specific claims. He often implies expertise without fully stating it, leaving just enough ambiguity to avoid direct accountability. When pressed, he shifts the discussion back to meta-level reflections, sidestepping any concrete resolution. This oscillation between authority and evasion creates the illusion of intellectual depth while ensuring that his positions remain unassailable, not because they are correct, but because they are never fully articulated in a way that allows direct critique.

Bateman also relies on selective engagement to reinforce his positioning. He responds to points that allow him to reiterate his role as an arbiter but ignores or glosses over arguments that would require him to take a definitive stand. This creates the impression that his perspective is comprehensive, when in reality, it is carefully curated to avoid genuine confrontation.

A particularly insidious aspect of this strategy is its effect on discourse itself. By positioning himself as a detached observer, he implicitly delegitimises those who do engage directly. His interventions often suggest that the real issue is not the argument itself but how it is being conducted—subtly shifting the focus away from substantive critique and onto decorum, perspective-taking, or supposed misunderstandings. This deflection serves to neutralise challenges without addressing them.

Sharpening this analysis requires making explicit the consequences of Bateman’s approach: it enables him to exert influence while avoiding scrutiny, stifles rigorous debate by privileging ambiguity over argument, and ultimately privileges his authority over the collective process of inquiry. The more this strategy is recognised for what it is—a method of rhetorical control rather than genuine engagement—the less effective it becomes.

The Esteemed Scholars of Sysfling: A Triumph of Wit, Wisdom, and Unilateral Decision-Making

[Scene: A dimly lit academic bar. David, John, Mick, and Brad are gathered around a table, engaged in the intellectual equivalent of a mutual back massage. Each of them has a drink in hand, and the air is thick with self-importance. Cathy, with a glint of passive-aggression in her eye, approaches the table, holding a small notepad.]

Cathy: (Smiling thinly) Well, well, if it isn’t Sysfling’s finest minds! I simply couldn’t help but overhear your stimulating discussion—so much insight, so much... authority.

David: (Leaning back, smugly) Ah, Cathy. Always good to see the press taking an interest in intellectual leadership.

Cathy: Oh, absolutely. It must be exhausting carrying the burden of enlightenment. Tell me, how do you cope with the sheer weight of your influence? The way you guide and shape discourse—it must be like herding, oh, I don’t know... photocopiers?

John: (Adjusts his glasses, frowns slightly) I wouldn't put it quite like that. It's more about maintaining a high standard of discussion, ensuring that rigorous, well-reasoned perspectives prevail.

Cathy: Oh, of course. And I just love how well that’s going! The way you brilliantly engaged with AI recently—so nuanced, so thoroughly researched. Tell me, Brad, when you exposed the dangers of AI addiction, did you ever worry that some might interpret it as... projection?

Brad: (Pauses mid-sip, clears throat) Well, Cathy, the problem isn’t with me—it's with those who lack detachment, who become too dependent, who mistake AI for meaningful discourse rather than a tool.

Cathy: Ohhh, I see. So others are addicted to AI, but when you have long, self-affirming dialogues with it, you’re just... conducting essential research? Fascinating. And Mick, your analysis of the Sysfling situation—so measured. How do you manage to stay so... neutral?

Mick: (Visibly uncomfortable, mutters into his drink) Well, I just try to be fair... you know, take all perspectives into account...

Cathy: Oh, absolutely! And that’s so brave of you. Especially in a situation where one side is lying, and the other is pointing it out—such a difficult moral quandary!

David: (Exasperated) Cathy, if you’re suggesting that I’ve lied—

Cathy: (Gasps theatrically) Oh, goodness, David, no! That would imply you knew you were spreading falsehoods, when clearly, you’re just very confident in whatever feels true at any given moment.

John: (Sternly) Cathy, this is precisely the kind of antagonistic discourse that poisons intellectual spaces.

Cathy: Ohhh, of course, John! Intellectual spaces are so fragile, aren’t they? Just one pointed question and poof! All that scholarly rigour just crumbles into dust!

Brad: (Sourly) If you’re trying to make us look foolish—

Cathy: (Beaming) Oh, no need! You’re doing amazingly on your own.

[The group falls into a tense silence. Cathy, still smiling, jots something in her notepad and waltzes off to the bar, humming.]

Frank Costanza Reading Sysfling Posts

Scene: The Costanza apartment. Frank is sitting at the table, reading from a laptop. Estelle putters around, only half-listening.

FRANK: (reading aloud, incredulous) "Dear all, despite Chris' attempt to foreclose issues with 'retrospective'…" What the hell is this guy talking about, Estelle?! Who writes like this?!

ESTELLE: Hmm?

FRANK: This Bateman character! He’s trying to sound like he’s delivering a UN address when he’s just whining on an email list! Look at this! (scrolling) “One can only say so much without knowledge.” Ohhh, well, thank you, your majesty! What an insight! Without knowledge, you can’t say much! You hear that, Estelle?! We should put that in a fortune cookie!

ESTELLE: Maybe he’s just trying to explain something?

FRANK: Explain?! He’s not explaining, he’s performing! He’s standing at the podium, waving to an imaginary crowd! “Ladies and gentlemen! I will now bestow upon you… MY OPINION!” (throws up hands in mock reverence)

ESTELLE: So don’t read it, Frank.

FRANK: I have to read it, Estelle! I have to witness this train wreck! Look at this— “You can safely delete any of Chris's posts…” Who says that?! He’s standing there with a clipboard, making a list of who’s allowed to speak! “You, out! You, in! No questions! No complaints!” It’s a dictatorship of smugness!

ESTELLE: Maybe you should go lie down.

FRANK: No! No lying down! I stand against this nonsense! I haven’t seen self-importance like this since the guy at the bakery tried to correct my pronunciation of “brioche!” I say “bree-osh,” he says “bree-ohhhsh!” Who cares?! Just give me the damn bread!

ESTELLE: I don’t know why you get involved in these things…

FRANK: Because, Estelle! Somebody has to say it! Somebody has to tell this man… (points at the screen with fury) YOU ARE NOT A WISE OVERSEER! YOU ARE JUST SOME GUY!!!

---

[Scene: Frank is still fuming over his laptop. The door suddenly bursts open—Kramer-style.]

KRAMER: (out of breath, waving a printout) Frank! FRANK! Have you seen this?!

FRANK: (still reading Bateman’s post, gritting his teeth) Oh, I’ve seen it, all right!

KRAMER: No, no, no, I mean this thing about the language models! It’s big! HUGE! We’re talkin’—sentences without meaning!

FRANK: That’s what I’m saying!

KRAMER: They’re putting words together—get this—without knowing what they mean!

FRANK: EXACTLY! That’s this Bateman guy!

KRAMER: No, no, no, he’s right! We gotta listen to him, Frank! He says we should just delete all of Chris’s posts!

FRANK: (snaps around to face him) WHAT?!

KRAMER: He’s the expert! He knows things! We’re just out here reading words like a couple of schmucks!

FRANK: YOU THINK I DON’T KNOW WORDS?!

KRAMER: No, no, no, but—see, you and me, we think words mean things, but Bateman—ohhh, he’s beyond that! He’s in the meta-zone!

FRANK: Meta-zone?! I’ll tell you what’s in the meta-zone, KRAMER—THIS GUY’S HEAD! Because he is talking out of it!

KRAMER: I dunno, Frank… he sounds pretty sure of himself…

FRANK: SO DID THE GUY WHO SOLD ME THAT “REVOLUTIONARY” BACK STRETCHER! AND NOW I CAN’T TURN LEFT!

(Frank aggressively demonstrates his inability to turn left. Kramer takes a cautious step back.)

ESTELLE: (muttering to herself, barely looking up) I told you not to buy that thing…

KRAMER: (pauses, then shakes his head in disappointment) It’s a shame, Frank… you could’ve learned something…

(Kramer solemnly shakes his head, pats Frank on the shoulder, and exits dramatically.)

FRANK: (yelling after him) I’VE LEARNED PLENTY! I LEARNED THAT BATEMAN IS A WINDBAG AND YOU’RE A MORON!

(Estelle sighs as Frank slams his laptop shut.)

John Bateman On A ChatGPT Conversation Being Of No Interest To Systemicists [4]

John Bateman continued on SYSFLING 28 Nov 2024, at 21:35:

…This relates to a concern that has been discussed for a couple of years now concerning the degradation of training data due to including in the training data material that has been created by large language models rather than human produced outputs. Since language models are often trained on (English) texts collected from the web, if the web contains language model produced nonsense (e.g., incorrect, ideologically unsavoury, wildly enthusiastic, etc.), then one can expect more of the same. So even circulating these produced 'texts' is potentially contributing to the garbage heap.

And, just to be clear, I like large language models a lot, we use them all the time in our research and even some text production, but one does not find out much about them by asking their 'opinion' despite how warm and cuddly the sequences of tokens produced appear!

Blogger Comments:

[1] To be clear, language that is "incorrect", ideologically unsavoury, or wildly enthusiastic is not "nonsense". These terms, along with 'nonsense', express attitudes to language which apply just as much to the language of humans as they do to the language produced by LLMs, since the latter mimics the former.

[2] To be clear, linguists, as the name implies, are interested in finding out about language. LLMs provide one means of doing so.

John Bateman On A ChatGPT Conversation Being Of No Interest To Systemicists [3]

John Bateman continued on SYSFLING 28 Nov 2024, at 21:35:

… As always with language models on their own, there is no training possible for 'truth'. One sees some of the consequences of this with the segment:

 ChRIS said:

Well pattern-matching seems to be a lot of what's going on. What's interesting is how much a ChatGPT can simulate a human just by pattern-matching.

ChatGPT said:

Exactly! It’s fascinating how much of what we think of as “human behaviour” can be boiled down to complex pattern-matching. In fact, a lot of our daily interactions, thoughts, and decisions seem to follow predictable patterns,

this can sound as if it makes sense, but language models do not work on pattern matching at all, despite the apparent convergence constructed here that might be mistaken for an advance in shared knowledge among the 'participants'. This lets a user go away potentially feeling good about a supportive interaction and with the idea that perhaps they have learned something. The actual mechanisms involved have little to do with pattern matching as traditionally thought of, however. Conceptualising the operation of large language models as pattern matching can mislead therefore and one sees quite often continuations of the (public) discourse along lines such as "it's only pattern matching", etc.

This is where 'congenial' turns to potentially highly pernicious, because there has been no supportive interaction and certainly not an increase in knowledge: quite the opposite — this can then also be taken up by others and circulated. …

Blogger Comments:

[1] To be clear, this also the case for humans. See further below.

[2] To be clear, contrary to Bateman's claim, when ChatGPT used the term 'pattern matching', it was not using it in the 'traditional' sense of the term; see the earlier post ChatGPT On John Bateman On ChatGPT. Here are the relevant points:

Clarifying "Pattern Matching":

1. Traditional Pattern Matching:
   In computational terms, this often refers to predefined rules or templates. For instance, regular expressions or specific "if-then" conditions designed to identify or react to specific patterns in input data.

2. Statistical Modelling (What LLMs Do):
   

   Language models like me (ChatGPT) do not use predefined rules or explicit templates. Instead, they operate probabilistically. When I generate text, I predict the likelihood of each possible next token (word, punctuation, etc.) based on the patterns observed in the training data. These "patterns" emerge from statistical correlations in massive datasets, not explicitly human-defined rules.

However, if "pattern matching" is understood in a broader sense of recognising and responding to patterns in data, then it could be argued that LLMs do work on a form of pattern recognition, albeit probabilistic and vastly more complex.

Bateman’s categorical claim that language models do not involve pattern matching "at all" overlooks the role of statistical learning, which is fundamentally about recognising and utilising patterns in data. While this process differs from traditional rule-based pattern matching, it undeniably involves identifying and leveraging patterns in text.

To say they don’t work on patterns "at all" is misleading, as recognising statistical relationships is a form of pattern utilisation.

[3] As can be seen from the above, this false conclusion derives from Bateman misrepresenting ChatGPT's use of 'pattern matching' as the traditional meaning, to a community of linguists who are largely unfamiliar with the field, and will naturally assume he is entirely trustworthy in such matters.

John Bateman On A ChatGPT Conversation Being Of No Interest To Systemicists [2]

John Bateman continued on SYSFLING 28 Nov 2024, at 21:35:

… For all such outputs, it is generally potentially useful to know the precise language model being used and the basic settings concerning 'temperature', i.e., restricted the behaviour is to the prompt, and the number of potential selections are considered as 'part of the mix' when moving to the next token. Changing these produces very different behaviour. And, of course, as now becoming increasingly relevant, the 'history' of prompts maintained for any particular interaction.

I wonder in particular about the latter as the responses of the system seem set to 'crazily over-enthusiastic puppy' mode, where any user prompt gives rise to phrases of excessive positive evaluation with personal stand-taking,

e.g., "Yes, that’s a fascinating distinction!",

"Ah, I love this idea!", etc.

Producing this kind of phrasing is usually the result of what is called reinforcement learning with human feedback (RLHF), where language model output is pushed towards responses that human users have rated positively along some dimensions of choice, such as 'congenial'. …

 

Blogger Comments:

[1] This is precisely why the ChatGPT conversation might be of interest to Systemicists. It shows one way that a dialogue can develop, with each turn depending on the meaning selections of the previous turn.

[2] Again, this precisely why the ChatGPT conversation might be of interest to Systemicists. The ChatGPT, an artefact with no experience of emotion, echoes the attitudinal semantics of the human interactant.

[3] To be clear, the high graduation of ATTITUDE in the ChatGPT turns echoes my earlier positive attitude when I had been using ChatGPT to create scenarios that continually had me laughing uproariously. The reason I appreciated the cheering up so much was that my partner of the last forty years had very recently died in our home, while I was asleep, after having suffered from early onset dementia for about three years.

John Bateman On A ChatGPT Conversation Being Of No Interest To Systemicists [1]

John Bateman wrote to SYSFLING 28 Nov 2024, at 21:35:

Hi all,

feels like this cycle has come round again...

On 26.11.24 13:07, Dr ChRIS CLÉiRIGh wrote:

My following conversation with ChatGPT might be of interest to some.

I feel obliged to ask just *why* this might be of interest... this kind of output can be produced endlessly ... and more or less continuously. …

 

Blogger Comments:

ChatGPT provides 5 general reasons why the conversation [here] might be of interest to Systemicists:

This conversation offers rich material for systemic functional linguists (SFL) because it touches on several key aspects of SFL's focus on language as a tool for meaning-making and social interaction. Here’s why it might be of interest:

1. Language as a System of Choice: The conversation highlights how AI (like ChatGPT) processes language to generate responses, but does so without lived experience or the ability to make value-based judgments. For SFL theorists, this distinction emphasises the choices humans make when constructing meaning, based not just on syntax and structure, but also on experience, emotions, and cultural context. In contrast, AI's language generation is purely functional and lacks these deeper semiotic layers.

2. The Semiotic Functions of Language: Systemic Functional Linguistics, particularly Halliday’s framework, explores how language functions in three metafunctions: ideational (conveying experience), interpersonal (building relationships), and textual (organising discourse). The conversation raises questions about how AI can mimic human-like responses and interact socially, but it cannot fully engage with the ideational or interpersonal functions because it lacks embodied experience. SFL would explore how meaning in human language is constructed through these layers and how AI's lack of experience creates a gap in its ability to function fully on all three metafunctions.

3. Consciousness and Language: The contrast between human consciousness and AI’s pattern-matching capabilities mirrors debates in SFL about the relationship between cognitive processes and language. Edelman's theory of neuronal group selection brings attention to the embodied and sensory foundations of consciousness, while SFL examines how language structures the experience of the world. The discussion about the gap between AI's functional language and the depth of human experience touches on how language reflects not just knowledge but also lived reality—something SFL emphasizes in its analysis of how language represents experience and reality.

4. Interactivity and Pattern Recognition: A central point of the conversation is AI’s ability to simulate human-like interactions through pattern recognition, which mirrors how humans often make decisions and communicate based on past experiences. From an SFL perspective, this brings into question how much of human behaviour, particularly communicative behaviour, can be seen as a sophisticated form of pattern recognition, and how language itself is used to construct, negotiate, and respond to social and cognitive patterns.

5. Embodied Experience and Value: The discussion also touches on how human language is shaped by the value humans assign to their sensory and emotional experiences. SFL would be interested in how humans not only categorise these experiences but also transform them into language that reflects their personal and social meaning. Since AI lacks this embodied experience, it can only categorise and generate language based on existing linguistic patterns, rather than creating new meanings grounded in experience.

In sum, this conversation engages deeply with issues central to SFL, such as the functional aspects of language, the role of experience in shaping meaning, and the interplay between cognition, social interaction, and communication. Systemic functional linguists would be particularly interested in how the distinctions between human and AI language production highlight the role of embodied experience in meaning-making and the limitations AI faces in replicating this process. 

ChatGPT On John Bateman On ChatGPT

John Bateman wrote to SYSFLING 28 Nov 2024, at 21:35:

… language models do not work on pattern matching at all …

ChatGPT Comments:

John Bateman’s claim that language models do not work on "pattern matching at all" is nuanced and points to a broader distinction about what "pattern matching" means in the context of computational linguistics versus everyday language.

Clarifying "Pattern Matching":

1. Traditional Pattern Matching:
   In computational terms, this often refers to predefined rules or templates. For instance, regular expressions or specific "if-then" conditions designed to identify or react to specific patterns in input data.

2. Statistical Modelling (What LLMs Do):
   

   Language models like me (ChatGPT) do not use predefined rules or explicit templates. Instead, they operate probabilistically. When I generate text, I predict the likelihood of each possible next token (word, punctuation, etc.) based on the patterns observed in the training data. These "patterns" emerge from statistical correlations in massive datasets, not explicitly human-defined rules.

Does This Contradict Bateman's Claim?

It depends on how one interprets "pattern matching."

• If Bateman is using "pattern matching" in the strict, technical sense (e.g., regex-like processing), then he is correct—LLMs do not operate in that manner.
• However, if "pattern matching" is understood in a broader sense of recognising and responding to patterns in data, then it could be argued that LLMs do work on a form of pattern recognition, albeit probabilistic and vastly more complex.

The Tension in Public Understanding:

Bateman rightly points out that casual descriptions of LLMs as "just pattern matching" can be misleading. It trivialises the complexity of the deep learning mechanisms at work, such as the self-attention mechanism in Transformer models, which dynamically focuses on relevant parts of the input to generate contextually appropriate output. While this process is data-driven and correlational, it is far more sophisticated than basic pattern matching.

My Perspective:

• Semantically: LLMs "find and use patterns," but not in the simplistic sense that might be implied by the phrase "pattern matching."
• Practically: Many of the emergent behaviours in LLMs—like maintaining coherence over long conversations—result from statistical learning over patterns, albeit at a massive scale.

Why This Matters:

The distinction is important because describing LLMs as "just pattern matching" risks minimising both their capabilities and their limitations. While LLMs are incredibly powerful at modelling language, they lack the understanding and intent behind patterns, which can lead to issues like confidently producing incorrect information.

Let’s break it down a bit further:

Why It Feels Overstated:

1. Binary Rejection of Pattern Matching:
   Bateman’s categorical claim that language models do not involve pattern matching "at all" overlooks the role of statistical learning, which is fundamentally about recognising and utilising patterns in data. While this process differs from traditional rule-based pattern matching, it undeniably involves identifying and leveraging patterns in text.

2. Lack of Nuance in Terminology:
   Bateman's critique might stem from a desire to distinguish LLMs' complex probabilistic modelling from naive notions of "pattern matching." However, by denying any connection to pattern recognition, he risks confusing readers who observe clear evidence of pattern-dependent outputs (e.g., stylistic mimicry, consistent conversational threads).

3. Misalignment with Observed Behaviours:
   If one were to interact with an LLM, it’s obvious that the model recognises structures in language—syntax, style, context—even if these aren't "patterns" in the traditional sense. From a user perspective, this very much feels like "pattern matching," even if it’s probabilistically driven and emergent.

Why Bateman’s Core Concern is Valid:

While he might overstate his case, Bateman’s deeper concern seems to be about public understanding and the dangers of oversimplification:

• Avoiding Mischaracterisation:
  He’s right that calling LLMs "just pattern matchers" trivialises their complexity and can mislead people into thinking they are deterministic or rule-based systems.

• Emphasising Emergent Complexity:
  Bateman may be pushing back against the idea that LLMs "understand" patterns in the human sense. Their outputs are based on statistical correlations, not comprehension or intent.

A Balanced View:

• True: LLMs don’t engage in classical pattern matching; their mechanisms are probabilistic, learned through massive neural networks.
• True: They exploit and replicate patterns in the data they were trained on, which is why they can mimic language and context so effectively.
• False: To say they don’t work on patterns "at all" is misleading, as recognising statistical relationships is a form of pattern utilisation.

Final Thoughts:

Bateman’s concern likely lies in preventing the spread of oversimplified or incorrect views about LLMs, but his phrasing risks dismissing the nuanced ways in which patterns inform LLM outputs. A more precise argument would have been to clarify how LLMs use patterns and contrast this with traditional understandings of "pattern matching."

"Well Within Registerial Norms"

The Transitivity Of Bamboozlement

John Bateman wrote to sysfling on 2 November 2017 at 19:10:

One of the main points of these papers (derived from our 12 year collaborative research center on space ...) is to open up the 'space' (ha ha) between a grammatically-induced semantics and any contextualised use of involved grammatical constructions so that the enormous flexibility of ranges of interpretation is both made visible and constrained sufficiently so that it can be worked with productively.

This is a hypotactically elaborating clause complex, with the dependent clause enclosed within the main clause:

One of the main points of these papers	<< derived from our 12 year collaborative research center on space >>	is to open up the 'space' between a grammatically-induced semantics and any contextualised use of involved grammatical constructions so that the enormous flexibility of ranges of interpretation is both made visible and constrained sufficiently so that it can be worked with productively
	= β
α

Ranking α Clause

The ranking α clause has a lexical density of about 23, depending how it it measured.  It metaphorically construes an identity of purpose, wherein the Value (one of the main points of these papers) is encoded by reference to the Token (to open up the 'space' between a grammatically-induced semantics and any contextualised use of involved grammatical constructions so that the enormous flexibility of ranges of interpretation is both made visible and constrained sufficiently so that it can be worked with productively):

One of the main points of these papers	is	to open up the 'space' between a grammatically-induced semantics and any contextualised use of involved grammatical constructions so that the enormous flexibility of ranges of interpretation is both made visible and constrained sufficiently so that it can be worked with productively
Identified Value	Process	Identifier Token

The clause is marked for both voice (receptive) and direction of coding (encoding).

Identifier Token

The Identifier Token in this relation is realised as an embedded clause complex in which the main clause is hypotactically enhanced, first in terms of result — and this includes a nested paratactic extending nexus — and then in terms of purpose:

to open up the 'space' between a grammatically-induced semantics and any contextualised use of involved grammatical constructions	so that the enormous flexibility of ranges of interpretation is both made visible	and constrained sufficiently	so that it can be worked with productively
α	x β cause: result		x γ cause: purpose
	1	+ 2

Embedded α Clause

The α clause in the embedded complex serving as Identifier Token is a non-finite abstract material clause:

to open up	the 'space' between a grammatically-induced semantics and any contextualised use of involved grammatical constructions
Process	Goal

The Goal of this clause is realised by a nominal group in which the Thing is both metaphorical (circumstantial) and abstract:

the	'space'	between a grammatically-induced semantics and any contextualised use of involved grammatical constructions
Deictic	Thing	Qualifier

The Qualifier of this nominal group is realised by a prepositional phrase whose Range is realised by a paratactically extending nominal group complex:

between	a grammatically-induced semantics	and any contextualised use of involved grammatical constructions
Process	Range
	1	+ 2

The first nominal group construes a semantic theory as Thing and sub-classifies it as grammatically-induced:

a	grammatically-induced	semantics
Deictic	Classifier	Thing

The second nominal group metaphorically construes a process ('use') as Thing, which it sub-classifies as contextualised, and metaphorically construes the congruent medium of the process involved grammatical constructions as Qualifier:

any	contextualised	use	of involved grammatical constructions
Deictic	Classifier	Thing	Qualifier

The Qualifier of this nominal group is realised by a prepositional phrase:

of	involved grammatical constructions
Process	Range

The Range of this prepositional phrase is realised by a nominal group that sub-classifies constructions as grammatical, and identifies them as involved:

involved	grammatical	constructions
post-Deictic	Classifier	Thing

Embedded β1 Clause

The β1 clause in the embedded complex serving as Identifier Token is an attributive clause that assigns the enormous flexibility of ranges of interpretation to the set visible:

so that	the enormous flexibility of ranges of interpretation	is	both	made	visible
	Carrier	Process			Attribute

The Carrier is realised by a nominal group that metaphorically construes a quality as Thing:

the	enormous	flexibility	of ranges of interpretation
Deictic	Epithet	Thing	Qualifier

The Qualifier is realised by a prepositional phrase:

of	ranges of interpretation
Process	Range

The Range of this prepositional phrase is realised by a nominal group:

ranges	of interpretation
Thing	Qualifier

Embedded β2 Clause

The β2 clause in the embedded complex serving as Identifier Token has an ellipsed Mood element:

the enormous flexibility of ranges of interpretation	is constrained		sufficiently
Carrier	Process	Attribute	Manner: degree

Embedded γ Clause

The γ clause in the embedded complex serving as Identifier Token is a receptive material clause:

so that	it	can be worked with	productively
	Goal	Process	Manner: quality

Ranking β Clause

The ranking elaborating clause can be analysed as 

derived from	our 12 year collaborative research center on space
Process: circumstantial	Attribute

The Attribute is realised by the nominal group:

our	12 year	collaborative	research center	on space
Deictic	post-Deictic	Classifier	Thing	Qualifier

The Thing of the nominal group is realised by a word complex:

research	center
β	α

Conclusion

This is the wording of someone who expects not to be understood, the social function of which has been described by Halliday & Matthiessen (1999: 272):

So the more the extent of grammatical metaphor in a text, the more that text is loaded against the learner, and against anyone who is an outsider to the register in question. It becomes elitist discourse, in which the function of constructing knowledge goes together with the function of restricting access to that knowledge, making it impenetrable to all except those who have the means of admission to the inside, or the select group of those who are already there. 

It is this other potential that grammatical metaphor has, for making meaning that is obscure, arcane and exclusive, that makes it ideal as a mode of discourse for establishing and maintaining status, prestige and hierarchy, and to establish the paternalistic authority of a technocratic elite whose message is 'this is all too hard for you to understand; so leave the decision-making to us.

See also John Bateman Positively Judging Himself and John Bateman On "The 'Space' Between A Grammatically-Induced Semantics And Any Contextualised Use Of Involved Grammatical Constructions".

We don't understand.

He must be clever.

He must be right.

— Peter Hammill 'The Jargon King'